Skip to contentSkip to conversation

dmaengine: dmatest: Fix regression when run command on misconfigured channel

PATCH In review Andy Shevchenko 3 participants 8 messages Sep 16, 2020 – Sep 22, 2020 dmaengine

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 dmaengine: dmatest: Fix regression when run command on misconfigured channel

From: Vladimir Murzin <[email protected]>

Andy reported that commit 6b41030fdc79 ("dmaengine: dmatest:
Restore default for channel") broke his scripts for the case where "busy" channel is used for configuration with expectation that run command would do nothing (and return 0). Instead, behavior was (unintentionally) changed to treat such case as under-configuration and progress with defaults, i.e. run command would start a test with default setting for channel (which would use all channels).

Restore original behavior with tracking status of channel setter so we can distinguish between misconfigured and under-configured cases in run command and act accordingly.

Fixes 6b41030fdc79 ("dmaengine: dmatest: Restore default for channel")
---
 drivers/dma/dmatest.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)
drivers/dma/dmatest.c +22 -6
index b2790641370a..4c9a9d7b48bb 100644
@@ -129,6 +129,7 @@ struct dmatest_params {
  * @nr_channels:	number of channels under test
  * @lock:		access protection to the fields of this structure
  * @did_init:		module has been initialized completely
+ * @last_error:		test has faced configuration issues
  */
 static struct dmatest_info {
 	/* Test parameters */
@@ -137,6 +138,7 @@ static struct dmatest_info {
 	/* Internal state */
 	struct list_head	channels;
 	unsigned int		nr_channels;
+	int			last_error;
 	struct mutex		lock;
 	bool			did_init;
 } test_info = {
@@ -1202,10 +1204,22 @@ static int dmatest_run_set(const char *val, const struct kernel_param *kp)
 		return ret;
 	} else if (dmatest_run) {
 		if (!is_threaded_test_pending(info)) {
-			pr_info("No channels configured, continue with any\n");
-			if (!is_threaded_test_run(info))
-				stop_threaded_test(info);
-			add_threaded_test(info);
+			/*
+			 * We have nothing to run. This can be due to:
+			 */
+			ret = info->last_error;
+			if (ret) {
+				/* 1) Mis-configuration */
+				pr_warn("Channel misconfigured, can't continue\n");
+				mutex_unlock(&info->lock);
+				return ret;
+			} else {
+				/* 2) We rely on defaults */
+				pr_info("No channels configured, continue with any\n");
+				if (!is_threaded_test_run(info))
+					stop_threaded_test(info);
+				add_threaded_test(info);
+			}
 		}
 		start_threaded_tests(info);
 	} else {
@@ -1222,7 +1236,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 	struct dmatest_info *info = &test_info;
 	struct dmatest_chan *dtc;
 	char chan_reset_val[20];
-	int ret = 0;
+	int ret;
 
 	mutex_lock(&info->lock);
 	ret = param_set_copystring(val, kp);
@@ -1230,7 +1244,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		mutex_unlock(&info->lock);
 		return ret;
 	}
-	/*Clear any previously run threads */
+	/* Clear any previously run threads */
 	if (!is_threaded_test_run(info) && !is_threaded_test_pending(info))
 		stop_threaded_test(info);
 	/* Reject channels that are already registered */
@@ -1277,12 +1291,14 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		goto add_chan_err;
 	}
 
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
 
 add_chan_err:
 	param_set_copystring(chan_reset_val, kp);
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
Signature · 1 line
2.28.0
#1 lore raw

PATCH 2/3 dmaengine: dmatest: Check list for emptiness before access its last entry

After writing a garbage to the channel we get an Oops in dmatest_chan_set() due to access to last entry in the empty list.

[  212.670672] BUG: unable to handle page fault for address: fffffff000000020
[  212.677562] #PF: supervisor read access in kernel mode
[  212.682702] #PF: error_code(0x0000) - not-present page
...
[  212.710074] RIP: 0010:dmatest_chan_set+0x149/0x2d0 [dmatest]
[  212.715739] Code: e8 cc f9 ff ff 48 8b 1d 0d 55 00 00 48 83 7b 10 00 0f 84 63 01 00 00 48 c7 c7 d0 65 4d c0 e8 ee 4a f5 e1 48 89 c6 48 8b 43 10 <48> 8b 40 20 48 8b 78 58 48 85 ff 0f 84 f5 00 00 00 e8 b1 41 f5 e1

Fix this by checking list for emptiness before accessing its last entry.

Fixes d53513d5dc28 ("dmaengine: dmatest: Add support for multi channel testing")
Cc Vladimir Murzin <[email protected]>
---
 drivers/dma/dmatest.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
drivers/dma/dmatest.c +2 -3
index 4c9a9d7b48bb..757eb1727a04 100644
@@ -1267,15 +1267,14 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 	add_threaded_test(info);
 
 	/* Check if channel was added successfully */
-	dtc = list_last_entry(&info->channels, struct dmatest_chan, node);
-
-	if (dtc->chan) {
+	if (!list_empty(&info->channels)) {
 		/*
 		 * if new channel was not successfully added, revert the
 		 * "test_channel" string to the name of the last successfully
 		 * added channel. exception for when users issues empty string
 		 * to channel parameter.
 		 */
+		dtc = list_last_entry(&info->channels, struct dmatest_chan, node);
 		if ((strcmp(dma_chan_name(dtc->chan), strim(test_channel)) != 0)
 		    && (strcmp("", strim(test_channel)) != 0)) {
 			ret = -EINVAL;
Signature · 1 line
2.28.0
#2 lore raw

PATCH 3/3 dmaengine: dmatest: Return boolean result directly in filter()

In reply to Andy Shevchenko

There is no need to have a conditional for boolean expression when function returns bool. Drop unnecessary code and return boolean result directly.

While at it, drop unneeded casting from void *.

---
 drivers/dma/dmatest.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)
drivers/dma/dmatest.c +1 -7
index 757eb1727a04..cf1379189316 100644
@@ -1070,13 +1070,7 @@ static int dmatest_add_channel(struct dmatest_info *info,
 
 static bool filter(struct dma_chan *chan, void *param)
 {
-	struct dmatest_params *params = param;
-
-	if (!dmatest_match_channel(params, chan) ||
-	    !dmatest_match_device(params, chan->device))
-		return false;
-	else
-		return true;
+	return dmatest_match_channel(param, chan) && dmatest_match_device(param, chan->device);
 }
 
 static void request_channels(struct dmatest_info *info,
Signature · 1 line
2.28.0
Tested-by on v1 3/3
#3 lore raw

Re: [PATCH v1 3/3] dmaengine: dmatest: Return boolean result directly in filter()

Hi Andy,

Quoted 5 lines · Andy Shevchenko

There is no need to have a conditional for boolean expression when function returns bool. Drop unnecessary code and return boolean result directly.

While at it, drop unneeded casting from void *.

my test scripts are working fine with the three patch applied.

Vinod: for all three patches ^^

Andy, Vladimir: thanks for the fixes!

- Péter

Quoted 16 lines
---
 drivers/dma/dmatest.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)
drivers/dma/dmatest.c +0 -1
index 757eb1727a04..cf1379189316 100644
@@ -1070,13 +1070,7 @@ static int dmatest_add_channel(struct dmatest_info *info,
 
 static bool filter(struct dma_chan *chan, void *param)
 {
-	struct dmatest_params *params = param;

drivers/dma/dmatest.c static int dmatest_add_channel(struct dmatest_info *info, · line 1076

Quoted 10 lines
drivers/dma/dmatest.c +1 -6
-
-	if (!dmatest_match_channel(params, chan) ||
-	    !dmatest_match_device(params, chan->device))
-		return false;
-	else
-		return true;
+	return dmatest_match_channel(param, chan) && dmatest_match_device(param, chan->device);
 }
 
 static void request_channels(struct dmatest_info *info,

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki

#4 lore raw

Re: [PATCH v1 3/3] dmaengine: dmatest: Return boolean result directly in filter()

Quoted 7 lines · Peter Ujfalusi
Quoted 5 lines · Andy Shevchenko

There is no need to have a conditional for boolean expression when function returns bool. Drop unnecessary code and return boolean result directly.

While at it, drop unneeded casting from void *.

Quoted 7 lines

my test scripts are working fine with the three patch applied.

Vinod: for all three patches ^^

Andy, Vladimir: thanks for the fixes!

Thanks for testing! I'll apply your tag to v2 (assuming you are okay with slight changes Vinod requested).

Signature · 2 lines
With Best Regards,
Andy Shevchenko
maintainer
#5 lore raw
In reply to Andy Shevchenko

HI Andy/Vladimir

Quoted 1 line · Andy Shevchenko

From: Vladimir Murzin <[email protected]>

Subject of the patch is not apt. It describes the effect of this patch, which is to fix the regression but does not describe the change. Please revise it to describe the change done in this patch

Quoted 8 lines

Andy reported that commit 6b41030fdc79 ("dmaengine: dmatest:
Restore default for channel") broke his scripts for the case where "busy" channel is used for configuration with expectation that run command would do nothing (and return 0). Instead, behavior was (unintentionally) changed to treat such case as under-configuration and progress with defaults, i.e. run command would start a test with default setting for channel (which would use all channels).

but a mis-configured channel returning success and doing nothing does not look as a good behaviour, I agree it broke Andy's script but the behaviour was not good to start with ;)

Quoted 38 lines

Restore original behavior with tracking status of channel setter so we can distinguish between misconfigured and under-configured cases in run command and act accordingly.

Fixes 6b41030fdc79 ("dmaengine: dmatest: Restore default for channel")
---
 drivers/dma/dmatest.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)
drivers/dma/dmatest.c +2 -1
index b2790641370a..4c9a9d7b48bb 100644
@@ -129,6 +129,7 @@ struct dmatest_params {
  * @nr_channels:	number of channels under test
  * @lock:		access protection to the fields of this structure
  * @did_init:		module has been initialized completely
+ * @last_error:		test has faced configuration issues
  */
 static struct dmatest_info {
 	/* Test parameters */
@@ -137,6 +138,7 @@ static struct dmatest_info {
 	/* Internal state */
 	struct list_head	channels;
 	unsigned int		nr_channels;
+	int			last_error;
 	struct mutex		lock;
 	bool			did_init;
 } test_info = {
@@ -1202,10 +1204,22 @@ static int dmatest_run_set(const char *val, const struct kernel_param *kp)
 		return ret;
 	} else if (dmatest_run) {
 		if (!is_threaded_test_pending(info)) {
-			pr_info("No channels configured, continue with any\n");

drivers/dma/dmatest.c static int dmatest_run_set(const char *val, const struct kernel_param *kp) · line 1213

Quoted 10 lines
drivers/dma/dmatest.c +7 -3
-			if (!is_threaded_test_run(info))
-				stop_threaded_test(info);
-			add_threaded_test(info);
+			/*
+			 * We have nothing to run. This can be due to:
+			 */
+			ret = info->last_error;
+			if (ret) {
+				/* 1) Mis-configuration */
+				pr_warn("Channel misconfigured, can't continue\n");

I would think this should be error

Quoted 13 lines
drivers/dma/dmatest.c +9 -0
+				mutex_unlock(&info->lock);
+				return ret;
+			} else {
+				/* 2) We rely on defaults */
+				pr_info("No channels configured, continue with any\n");
+				if (!is_threaded_test_run(info))
+					stop_threaded_test(info);
+				add_threaded_test(info);
+			}
 		}
 		start_threaded_tests(info);
 	} else {

drivers/dma/dmatest.c static int dmatest_chan_set(const char *val, const struct kernel_param *kp) · line 1239

Quoted 5 lines
drivers/dma/dmatest.c +1 -1
@@ -1222,7 +1236,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 	struct dmatest_info *info = &test_info;
 	struct dmatest_chan *dtc;
 	char chan_reset_val[20];
-	int ret = 0;
+	int ret;

Does this belong to this fix?

drivers/dma/dmatest.c static int dmatest_chan_set(const char *val, const struct kernel_param *kp) · line 1247

Quoted 9 lines
drivers/dma/dmatest.c +1 -1
 	mutex_lock(&info->lock);
 	ret = param_set_copystring(val, kp);
@@ -1230,7 +1244,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		mutex_unlock(&info->lock);
 		return ret;
 	}
-	/*Clear any previously run threads */
+	/* Clear any previously run threads */

Lets keep style issues away from fixes please

Quoted 20 lines
drivers/dma/dmatest.c +2 -0
 	if (!is_threaded_test_run(info) && !is_threaded_test_pending(info))
 		stop_threaded_test(info);
 	/* Reject channels that are already registered */
@@ -1277,12 +1291,14 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		goto add_chan_err;
 	}
 
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
 
 add_chan_err:
 	param_set_copystring(chan_reset_val, kp);
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
Signature · 1 line
2.28.0
Signature · 1 line
~Vinod
#6 lore raw

On Thu, Sep 17, 2020 at 04:00:34PM +0530, Vinod Koul wrote:

Thanks for review, my answers below.

Quoted 6 lines · Andy Shevchenko
Quoted 1 line

From: Vladimir Murzin <[email protected]>

Subject of the patch is not apt. It describes the effect of this patch, which is to fix the regression but does not describe the change. Please revise it to describe the change done in this patch

OK!

Quoted 9 lines
Quoted 8 lines

Andy reported that commit 6b41030fdc79 ("dmaengine: dmatest:
Restore default for channel") broke his scripts for the case where "busy" channel is used for configuration with expectation that run command would do nothing (and return 0). Instead, behavior was (unintentionally) changed to treat such case as under-configuration and progress with defaults, i.e. run command would start a test with default setting for channel (which would use all channels).

Quoted 3 lines

but a mis-configured channel returning success and doing nothing does not look as a good behaviour, I agree it broke Andy's script but the behaviour was not good to start with ;)

Which used to be a previous behaviour. I don't understand what should I do here as after this patch (and even after the initial multi-channel support patch) the behaviour is like you desire.

Quoted 42 lines
Quoted 42 lines

Restore original behavior with tracking status of channel setter so we can distinguish between misconfigured and under-configured cases in run command and act accordingly.

Fixes 6b41030fdc79 ("dmaengine: dmatest: Restore default for channel")
---
 drivers/dma/dmatest.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)
drivers/dma/dmatest.c +3 -4
index b2790641370a..4c9a9d7b48bb 100644
@@ -129,6 +129,7 @@ struct dmatest_params {
  * @nr_channels:	number of channels under test
  * @lock:		access protection to the fields of this structure
  * @did_init:		module has been initialized completely
+ * @last_error:		test has faced configuration issues
  */
 static struct dmatest_info {
 	/* Test parameters */
@@ -137,6 +138,7 @@ static struct dmatest_info {
 	/* Internal state */
 	struct list_head	channels;
 	unsigned int		nr_channels;
+	int			last_error;
 	struct mutex		lock;
 	bool			did_init;
 } test_info = {
@@ -1202,10 +1204,22 @@ static int dmatest_run_set(const char *val, const struct kernel_param *kp)
 		return ret;
 	} else if (dmatest_run) {
 		if (!is_threaded_test_pending(info)) {
-			pr_info("No channels configured, continue with any\n");
-			if (!is_threaded_test_run(info))
-				stop_threaded_test(info);
-			add_threaded_test(info);
+			/*
Quoted 8 lines

drivers/dma/dmatest.c static int dmatest_run_set(const char *val, const struct kernel_param *kp) · line 1213

Quoted 6 lines
drivers/dma/dmatest.c +6 -0
+			 * We have nothing to run. This can be due to:
+			 */
+			ret = info->last_error;
+			if (ret) {
+				/* 1) Mis-configuration */
+				pr_warn("Channel misconfigured, can't continue\n");

I would think this should be error

OK!

Quoted 12 lines
Quoted 12 lines
drivers/dma/dmatest.c +9 -0
+				mutex_unlock(&info->lock);
+				return ret;
+			} else {
+				/* 2) We rely on defaults */
+				pr_info("No channels configured, continue with any\n");
+				if (!is_threaded_test_run(info))
+					stop_threaded_test(info);
+				add_threaded_test(info);
+			}
 		}
 		start_threaded_tests(info);
 	} else {
Quoted 8 lines

drivers/dma/dmatest.c static int dmatest_chan_set(const char *val, const struct kernel_param *kp) · line 1239

Quoted 6 lines
drivers/dma/dmatest.c +1 -1
@@ -1222,7 +1236,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 	struct dmatest_info *info = &test_info;
 	struct dmatest_chan *dtc;
 	char chan_reset_val[20];
-	int ret = 0;
+	int ret;

Does this belong to this fix?

Relatively. It goes under category that we can clean this up because it has relation to what we need below. I will leave it in v2 as it is in v1.

Quoted 10 lines

drivers/dma/dmatest.c static int dmatest_chan_set(const char *val, const struct kernel_param *kp) · line 1247

Quoted 8 lines
drivers/dma/dmatest.c +1 -1
 	mutex_lock(&info->lock);
 	ret = param_set_copystring(val, kp);
@@ -1230,7 +1244,7 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		mutex_unlock(&info->lock);
 		return ret;
 	}
-	/*Clear any previously run threads */
+	/* Clear any previously run threads */

Lets keep style issues away from fixes please

OK!

Quoted 23 lines
Quoted 20 lines
drivers/dma/dmatest.c +2 -0
 	if (!is_threaded_test_run(info) && !is_threaded_test_pending(info))
 		stop_threaded_test(info);
 	/* Reject channels that are already registered */
@@ -1277,12 +1291,14 @@ static int dmatest_chan_set(const char *val, const struct kernel_param *kp)
 		goto add_chan_err;
 	}
 
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
 
 add_chan_err:
 	param_set_copystring(chan_reset_val, kp);
+	info->last_error = ret;
 	mutex_unlock(&info->lock);
 
 	return ret;
Signature · 1 line
2.28.0
Signature · 1 line
~Vinod
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#7 lore raw
Quoted 13 lines · Andy Shevchenko
Quoted 12 lines · Vinod Koul
Quoted 10 lines · Andy Shevchenko

From: Vladimir Murzin <[email protected]>

Andy reported that commit 6b41030fdc79 ("dmaengine: dmatest:
Restore default for channel") broke his scripts for the case where "busy" channel is used for configuration with expectation that run command would do nothing (and return 0). Instead, behavior was (unintentionally) changed to treat such case as under-configuration and progress with defaults, i.e. run command would start a test with default setting for channel (which would use all channels).

Quoted 7 lines
Quoted 3 lines

but a mis-configured channel returning success and doing nothing does not look as a good behaviour, I agree it broke Andy's script but the behaviour was not good to start with ;)

Which used to be a previous behaviour. I don't understand what should I do here as after this patch (and even after the initial multi-channel support patch) the behaviour is like you desire.

Okay, I have dropped the part '(and return 0)' to avoid ambiguity.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Restore original behavior with tracking status of channel setter so we can distinguish between misconfigured and under-configured cases in run command and act accordingly.

Signature · 2 lines
With Best Regards,
Andy Shevchenko